[Clang] Allow additional mathematical symbols in identifiers.

[cor3ntin]

Summary

Implement the proposed UAX Profile
"Mathematical notation profile for default identifiers".

This implements a not-yet approved Unicode for a vetted
UAX31 identifier profile
https://www.unicode.org/L2/L2022/22230-math-profile.pdf

This change mitigates the reported disruption caused
by the implementation of UAX31 in C++ and C2x,
as these mathematical symbols are commonly used in the
scientific community.

Fixes #54732

[cor3ntin]

Note that we should wait end of next week to hear back from the Unicode Technical Committee before merging this

[cor3ntin]

Remove extra semi colon.

[cor3ntin]

@tahonermann gentle ping :)

[tahonermann]

I apologize for the delay Corentin, I hope to get through a number of my outstanding code reviews tomorrow.

[cor3ntin]

@tahonermann Gentle ping :)

[tahonermann]

This looks great to me. I added a minor comment regarding the grammar of the added diagnostic, but am accepting the review regardless.

clang/include/clang/Basic/DiagnosticLexKinds.td

129–130

I think this diagnostic should also have an "an" before "identifier" if you wish to take on correcting it as well.

135–137

The grammar seems slightly off here. I think there should be an "an" before "identifier":

• "mathematical notation character <U+%0> in an identifier is a Clang extension"

This would be consistent with err_character_not_allowed_identifier and warn_c99_compat_unicode_id.

[tahonermann]

Oh, also, the review summary mentions "UAX32" (twice); those should be "UAX31". Just please make sure the commit message is correct.

Also, I'm terribly sorry for being so slow to get this reviewed!

[cor3ntin]

In D137051#3982833, @tahonermann wrote:

Oh, also, the review summary mentions "UAX32" (twice); those should be "UAX31". Just please make sure the commit message is correct.

Also, I'm terribly sorry for being so slow to get this reviewed!

I'm glad you noticed that...!

[cor3ntin]

Fix grammar in diagnostic.

[cor3ntin]

Thanks Tom. I'm adding more reviewers in case someone else wants to look at it

[cor3ntin]

clang/include/clang/Basic/DiagnosticLexKinds.td

129–130

Done https://github.com/llvm/llvm-project/commit/9e2dc984ba1b25625e6ddc38ba681b194172379a

[cor3ntin]

clang-format

clang/docs/ReleaseNotes.rst

- When using header modules, inclusion of a private header and violations of
   <https://clang.llvm.org/docs/Modules.html#use-declaration>`_ are now
   diagnosed even when the includer is a textual header. This change can be
   temporarily reversed with ``-Xclang
   -fno-modules-validate-textual-header-includes``, but this flag will be
   removed in a future Clang release.
 - Unicode support has been updated to support Unicode 15.0.
   New unicode codepoints are supported as appropriate in diagnostics,
   C and C++ identifiers, and escape sequences.
+- In identifiers, Clang allows a restricted set of additional mathematical symbols
+  as an extension. These symbols correspond to a proposed Unicode
+  `Mathematical notation profile for default identifiers
+  <https://www.unicode.org/L2/L2022/22230-math-profile.pdf>`_.
+  This resolves `Issue 54732 <https://github.com/llvm/llvm-project/issues/54732>`_.
 - Clang now supports loading multiple configuration files. The files from
   default configuration paths are loaded first, unless ``--no-default-config``
   option is used. All files explicitly specified using ``--config=`` option
   are loaded afterwards.
 - When loading default configuration files, clang now unconditionally uses
   the real target triple (respecting options such as ``--target=`` and ``-m32``)
   rather than the executable prefix. The respective configuration files are
   also loaded when clang is called via an executable without a prefix (e.g.

clang/include/clang/Basic/DiagnosticLexKinds.td

def warn_invalid_utf8_in_comment : Extension<
   "invalid UTF-8 in comment">, InGroup<DiagGroup<"invalid-utf8">>;
 def err_character_not_allowed : Error<
   "unexpected character <U+%0>">;
 def err_character_not_allowed_identifier : Error<
   "character <U+%0> not allowed %select{in|at the start of}1 an identifier">;
 def ext_unicode_whitespace : ExtWarn<
   "treating Unicode character as whitespace">,
   InGroup<DiagGroup<"unicode-whitespace">>;
 def warn_utf8_symbol_homoglyph : Warning<
   "treating Unicode character <U+%0> as identifier character rather than "
   "as '%1' symbol">, InGroup<DiagGroup<"unicode-homoglyph">>;
 def warn_utf8_symbol_zero_width : Warning<
   "identifier contains Unicode character <U+%0> that is invisible in "
   "some environments">, InGroup<DiagGroup<"unicode-zero-width">>;
+def ext_mathematical_notation : ExtWarn<
+  "mathematical notation character <U+%0> in an identifier is a Clang extension">,
+  InGroup<DiagGroup<"mathematical-notation-identifier-extension">>;
 
 def ext_delimited_escape_sequence : Extension<
   "%select{delimited|named}0 escape sequences are a "
   "%select{Clang|C++2b}1 extension">,
   InGroup<DiagGroup<"delimited-escape-sequence-extension">>;
 
 def warn_cxx2b_delimited_escape_sequence : Warning<
   "%select{delimited|named}0 escape sequences are "

clang/lib/Lex/Lexer.cpp

 }
 
 static bool isUnicodeWhitespace(uint32_t Codepoint) {
   static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
       UnicodeWhitespaceCharRanges);
   return UnicodeWhitespaceChars.contains(Codepoint);
 }
 
-static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
+static llvm::SmallString<5> codepointAsHexString(uint32_t C) {
+  llvm::SmallString<5> CharBuf;
+  llvm::raw_svector_ostream CharOS(CharBuf);
+  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
+  return CharBuf;
+}
+
+// To mitigate https://github.com/llvm/llvm-project/issues/54732,
+// we allow "Mathematical Notation Characters" in identifiers.
+// This is a proposed profile that extends the XID_Start/XID_continue
+// with mathematical symbols, superscipts and subscripts digits
+// found in some production software.
+// https://www.unicode.org/L2/L2022/22230-math-profile.pdf
+static bool isMathematicalExtensionID(uint32_t C, const LangOptions &LangOpts, bool IsStart, bool & IsExtension) {
+  static const llvm::sys::UnicodeCharSet MathStartChars(MathematicalNotationProfileIDStartRanges);
+  static const llvm::sys::UnicodeCharSet MathContinueChars(MathematicalNotationProfileIDContinueRanges);
+  if(MathStartChars.contains(C) || (!IsStart && MathContinueChars.contains(C))) {
+    IsExtension = true;
+    return true;
+  }
+  return false;
+}
+
+static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts, bool & IsExtension) {
   if (LangOpts.AsmPreprocessor) {
     return false;
   } else if (LangOpts.DollarIdents && '$' == C) {
     return true;
   } else if (LangOpts.CPlusPlus || LangOpts.C2x) {
     // A non-leading codepoint must have the XID_Continue property.
     // XIDContinueRanges doesn't contains characters also in XIDStartRanges,
     // so we need to check both tables.
     // '_' doesn't have the XID_Continue property but is allowed in C and C++.
     static const llvm::sys::UnicodeCharSet XIDStartChars(XIDStartRanges);
     static const llvm::sys::UnicodeCharSet XIDContinueChars(XIDContinueRanges);
-    return C == '_' || XIDStartChars.contains(C) ||
-           XIDContinueChars.contains(C);
+    if(C == '_' || XIDStartChars.contains(C) ||
+        XIDContinueChars.contains(C))
+      return true;
+    return isMathematicalExtensionID(C, LangOpts, /*IsStart=*/false, IsExtension);
   } else if (LangOpts.C11) {
     static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
         C11AllowedIDCharRanges);
     return C11AllowedIDChars.contains(C);
   } else {
     static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
         C99AllowedIDCharRanges);
     return C99AllowedIDChars.contains(C);
   }
 }
 
-static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
+static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts, bool & IsExtension) {
   assert(C > 0x7F && "isAllowedInitiallyIDChar called with an ASCII codepoint");
+  IsExtension = false;
   if (LangOpts.AsmPreprocessor) {
     return false;
   }
   if (LangOpts.CPlusPlus || LangOpts.C2x) {
     static const llvm::sys::UnicodeCharSet XIDStartChars(XIDStartRanges);
-    return XIDStartChars.contains(C);
+    if(XIDStartChars.contains(C))
+      return true;
+    return isMathematicalExtensionID(C, LangOpts, /*IsStart=*/true, IsExtension);
   }
-  if (!isAllowedIDChar(C, LangOpts))
+  if (!isAllowedIDChar(C, LangOpts, IsExtension))
     return false;
   if (LangOpts.C11) {
     static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
         C11DisallowedInitialIDCharRanges);
     return !C11DisallowedInitialIDChars.contains(C);
   }
   static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
       C99DisallowedInitialIDCharRanges);
   return !C99DisallowedInitialIDChars.contains(C);
 }
 
+static void diagnoseExtensionInIdentifier(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range) {
+
+  static const llvm::sys::UnicodeCharSet MathStartChars(MathematicalNotationProfileIDStartRanges);
+  static const llvm::sys::UnicodeCharSet MathContinueChars(MathematicalNotationProfileIDContinueRanges);
+
+  assert((MathStartChars.contains(C) || MathContinueChars.contains(C))
+         && "Unexpected mathematical notation codepoint");
+  Diags.Report(Range.getBegin(), diag::ext_mathematical_notation)
+      << codepointAsHexString(C) << Range;
+}
+
+
 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
                                             const char *End) {
   return CharSourceRange::getCharRange(L.getSourceLocation(Begin),
                                        L.getSourceLocation(End));
 }
 
 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
                                       CharSourceRange Range, bool IsFirst) {
static constexpr HomoglyphPair SortedHomoglyphs[] = {
     {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
     {U'\uff5e', '~'}, // FULLWIDTH TILDE
     {0, 0}
   };
   auto Homoglyph =
       std::lower_bound(std::begin(SortedHomoglyphs),
                        std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
   if (Homoglyph->Character == C) {
-    llvm::SmallString<5> CharBuf;
-    {
-      llvm::raw_svector_ostream CharOS(CharBuf);
-      llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
-    }
     if (Homoglyph->LooksLike) {
       const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
       Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
-          << Range << CharBuf << LooksLikeStr;
+          << Range << codepointAsHexString(C) << LooksLikeStr;
     } else {
       Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
-          << Range << CharBuf;
+          << Range << codepointAsHexString(C);
     }
   }
 }
 
 static void diagnoseInvalidUnicodeCodepointInIdentifier(
     DiagnosticsEngine &Diags, const LangOptions &LangOpts, uint32_t CodePoint,
     CharSourceRange Range, bool IsFirst) {
   if (isASCII(CodePoint))
     return;
 
-  bool IsIDStart = isAllowedInitiallyIDChar(CodePoint, LangOpts);
-  bool IsIDContinue = IsIDStart || isAllowedIDChar(CodePoint, LangOpts);
+  bool IsExtension;
+  bool IsIDStart = isAllowedInitiallyIDChar(CodePoint, LangOpts, IsExtension);
+  bool IsIDContinue = IsIDStart || isAllowedIDChar(CodePoint, LangOpts, IsExtension);
 
   if ((IsFirst && IsIDStart) || (!IsFirst && IsIDContinue))
     return;
 
   bool InvalidOnlyAtStart = IsFirst && !IsIDStart && IsIDContinue;
 
-  llvm::SmallString<5> CharBuf;
-  llvm::raw_svector_ostream CharOS(CharBuf);
-  llvm::write_hex(CharOS, CodePoint, llvm::HexPrintStyle::Upper, 4);
-
   if (!IsFirst || InvalidOnlyAtStart) {
     Diags.Report(Range.getBegin(), diag::err_character_not_allowed_identifier)
-        << Range << CharBuf << int(InvalidOnlyAtStart)
+        << Range << codepointAsHexString(CodePoint) << int(InvalidOnlyAtStart)
         << FixItHint::CreateRemoval(Range);
   } else {
     Diags.Report(Range.getBegin(), diag::err_character_not_allowed)
-        << Range << CharBuf << FixItHint::CreateRemoval(Range);
+        << Range << codepointAsHexString(CodePoint) << FixItHint::CreateRemoval(Range);
   }
 }
 
 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
                                     Token &Result) {
   const char *UCNPtr = CurPtr + Size;
   uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
   if (CodePoint == 0) {
     return false;
   }
-
-  if (!isAllowedIDChar(CodePoint, LangOpts)) {
+  bool IsExtension = false;
+  if (!isAllowedIDChar(CodePoint, LangOpts, IsExtension)) {
     if (isASCII(CodePoint) || isUnicodeWhitespace(CodePoint))
       return false;
     if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
         !PP->isPreprocessedOutput())
       diagnoseInvalidUnicodeCodepointInIdentifier(
           PP->getDiagnostics(), LangOpts, CodePoint,
           makeCharRange(*this, CurPtr, UCNPtr),
           /*IsFirst=*/false);
 
     // We got a unicode codepoint that is neither a space nor a
     // a valid identifier part.
     // Carry on as if the codepoint was valid for recovery purposes.
-  } else if (!isLexingRawMode())
+  } else if (!isLexingRawMode()) {
+    if(IsExtension)
+      diagnoseExtensionInIdentifier(PP->getDiagnostics(), CodePoint,
+                                            makeCharRange(*this, CurPtr, UCNPtr));
+
     maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
                               makeCharRange(*this, CurPtr, UCNPtr),
                               /*IsFirst=*/false);
+  }
 
   Result.setFlag(Token::HasUCN);
   if ((UCNPtr - CurPtr ==  6 && CurPtr[1] == 'u') ||
       (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
     CurPtr = UCNPtr;
   else
     while (CurPtr != UCNPtr)
       (void)getAndAdvanceChar(CurPtr, Result);
   return true;
 }
 
 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
   const char *UnicodePtr = CurPtr;
   llvm::UTF32 CodePoint;
   llvm::ConversionResult Result =
       llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
                                 (const llvm::UTF8 *)BufferEnd,
                                 &CodePoint,
                                 llvm::strictConversion);
   if (Result != llvm::conversionOK)
     return false;
 
-  if (!isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts)) {
+  bool IsExtension = false;
+  if (!isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts, IsExtension)) {
     if (isASCII(CodePoint) || isUnicodeWhitespace(CodePoint))
       return false;
 
     if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
         !PP->isPreprocessedOutput())
       diagnoseInvalidUnicodeCodepointInIdentifier(
           PP->getDiagnostics(), LangOpts, CodePoint,
           makeCharRange(*this, CurPtr, UnicodePtr), /*IsFirst=*/false);
     // We got a unicode codepoint that is neither a space nor a
     // a valid identifier part. Carry on as if the codepoint was
     // valid for recovery purposes.
   } else if (!isLexingRawMode()) {
+    if(IsExtension)
+      diagnoseExtensionInIdentifier(PP->getDiagnostics(), CodePoint,
+                                            makeCharRange(*this, CurPtr, UnicodePtr));
     maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
                               makeCharRange(*this, CurPtr, UnicodePtr),
                               /*IsFirst=*/false);
     maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
                                makeCharRange(*this, CurPtr, UnicodePtr));
   }
 
   CurPtr = UnicodePtr;
   return true;
 }
 
 bool Lexer::LexUnicodeIdentifierStart(Token &Result, uint32_t C,
                                       const char *CurPtr) {
-  if (isAllowedInitiallyIDChar(C, LangOpts)) {
+  bool IsExtension = false;
+  if (isAllowedInitiallyIDChar(C, LangOpts, IsExtension)) {
     if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
         !PP->isPreprocessedOutput()) {
+      if(IsExtension)
+        diagnoseExtensionInIdentifier(PP->getDiagnostics(), C,
+                                              makeCharRange(*this, BufferPtr, CurPtr));
       maybeDiagnoseIDCharCompat(PP->getDiagnostics(), C,
                                 makeCharRange(*this, BufferPtr, CurPtr),
                                 /*IsFirst=*/true);
       maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), C,
                                  makeCharRange(*this, BufferPtr, CurPtr));
     }
 
     MIOpt.ReadToken();
     return LexIdentifierContinue(Result, CurPtr);
   }
 
   if (!isLexingRawMode() && !ParsingPreprocessorDirective &&
       !PP->isPreprocessedOutput() && !isASCII(*BufferPtr) &&
-      !isAllowedInitiallyIDChar(C, LangOpts) && !isUnicodeWhitespace(C)) {
+      !isUnicodeWhitespace(C)) {
     // Non-ASCII characters tend to creep into source code unintentionally.
     // Instead of letting the parser complain about the unknown token,
     // just drop the character.
     // Note that we can /only/ do this when the non-ASCII character is actually
     // spelled as Unicode, not written as a UCN. The standard requires that
     // we not throw away any possible preprocessor tokens, but there's a
     // loophole in the mapping of Unicode characters to basic character set
     // characters that allows us to map these particular characters to, say,

clang/lib/Lex/UnicodeCharSets.h

static const llvm::sys::UnicodeCharRange XIDContinueRanges[] = {
     {0x1DA3B, 0x1DA6C}, {0x1DA75, 0x1DA75}, {0x1DA84, 0x1DA84},
     {0x1DA9B, 0x1DA9F}, {0x1DAA1, 0x1DAAF}, {0x1E000, 0x1E006},
     {0x1E008, 0x1E018}, {0x1E01B, 0x1E021}, {0x1E023, 0x1E024},
     {0x1E026, 0x1E02A}, {0x1E08F, 0x1E08F}, {0x1E130, 0x1E136},
     {0x1E140, 0x1E149}, {0x1E2AE, 0x1E2AE}, {0x1E2EC, 0x1E2F9},
     {0x1E4EC, 0x1E4F9}, {0x1E8D0, 0x1E8D6}, {0x1E944, 0x1E94A},
     {0x1E950, 0x1E959}, {0x1FBF0, 0x1FBF9}, {0xE0100, 0xE01EF}};
 
+// Clang supports the "Mathematical notation profile" as an extension,
+// as described in https://www.unicode.org/L2/L2022/22230-math-profile.pdf
+// Math_Start
+static const llvm::sys::UnicodeCharRange MathematicalNotationProfileIDStartRanges[] = {
+    {0x02202, 0x02202}, // ∂
+    {0x02207, 0x02207}, // ∇
+    {0x0221E, 0x0221E}, // ∞
+    {0x1D6C1, 0x1D6C1}, // 𝛁
+    {0x1D6DB, 0x1D6DB}, // 𝛛
+    {0x1D6FB, 0x1D6FB}, // 𝛻
+    {0x1D715, 0x1D715}, // 𝜕
+    {0x1D735, 0x1D735}, // 𝜵
+    {0x1D74F, 0x1D74F}, // 𝝏
+    {0x1D76F, 0x1D76F}, // 𝝯
+    {0x1D789, 0x1D789}, // 𝞉
+    {0x1D7A9, 0x1D7A9}, // 𝞩
+    {0x1D7C3, 0x1D7C3}, // 𝟃
+};
+
+// Math_Continue
+static const llvm::sys::UnicodeCharRange MathematicalNotationProfileIDContinueRanges[] = {
+    {0x000B2, 0x000B3}, // ²-³
+    {0x000B9, 0x000B9}, // ¹
+    {0x02070, 0x02070}, // ⁰
+    {0x02074, 0x0207E}, // ⁴-⁾
+    {0x02080, 0x0208E}, // ₀-₎
+};
+
 // C11 D.1, C++11 [charname.allowed]
 static const llvm::sys::UnicodeCharRange C11AllowedIDCharRanges[] = {
   // 1
   { 0x00A8, 0x00A8 }, { 0x00AA, 0x00AA }, { 0x00AD, 0x00AD },
   { 0x00AF, 0x00AF }, { 0x00B2, 0x00B5 }, { 0x00B7, 0x00BA },
   { 0x00BC, 0x00BE }, { 0x00C0, 0x00D6 }, { 0x00D8, 0x00F6 },
   { 0x00F8, 0x00FF },
   // 2

clang/test/Driver/autocomplete.c

 // MRELOCMODELALL-NEXT: rwpi
 // MRELOCMODELALL-NEXT: static
 // RUN: %clang --autocomplete=-Wma | FileCheck %s -check-prefix=WARNING
 // WARNING: -Wmacro-redefined
 // WARNING-NEXT: -Wmain
 // WARNING-NEXT: -Wmain-return-type
 // WARNING-NEXT: -Wmalformed-warning-check
 // WARNING-NEXT: -Wmany-braces-around-scalar-init
+// WARNING-NEXT: -Wmathematical-notation-identifier-extension
 // WARNING-NEXT: -Wmax-tokens
 // WARNING-NEXT: -Wmax-unsigned-zero
 // RUN: %clang --autocomplete=-Wno-invalid-pp- | FileCheck %s -check-prefix=NOWARNING
 // NOWARNING: -Wno-invalid-pp-token
 // RUN: %clang --autocomplete=-analyzer-checker | FileCheck %s -check-prefix=ANALYZER
 // ANALYZER: unix.Malloc
 // RUN: %clang --autocomplete=-std= | FileCheck %s -check-prefix=STDVAL
 // STDVAL: c99

clang/test/Lexer/unicode.c

 extern int ꙮ;
 extern int  \u1B4C;     // BALINESE LETTER ARCHAIC JNYA - Added in Unicode 14
 extern int  \U00016AA2; // TANGSA LETTER GA - Added in Unicode 14
 extern int  \U0001E4D0; // 𞓐 NAG MUNDARI LETTER O - Added in Unicode 15
 extern int _\N{TANGSA LETTER GA};
 extern int _\N{TANGSALETTERGA}; // expected-error {{'TANGSALETTERGA' is not a valid Unicode character name}} \
                                 // expected-note {{characters names in Unicode escape sequences are sensitive to case and whitespace}}
 
+extern int 𝛛; // expected-warning {{mathematical notation character <U+1D6DB> in an identifier is a Clang extension}}
+extern int ₉; // expected-error {{character <U+2089> not allowed at the start of an identifier}} \\
+                 expected-warning {{declaration does not declare anything}}
+
+int a¹b₍₄₂₎∇; // expected-warning 6{{mathematical notation character}}
+
+int \u{221E} = 1; // expected-warning {{mathematical notation character}}
+int \N{MATHEMATICAL SANS-SERIF BOLD ITALIC PARTIAL DIFFERENTIAL} = 1;
+                 // expected-warning@-1 {{mathematical notation character}}
 
+int a\N{SUBSCRIPT EQUALS SIGN} = 1; // expected-warning {{mathematical notation character}}
 
 // This character doesn't have the XID_Start property
 extern int  \U00016AC0; // TANGSA DIGIT ZERO  // cxx-error {{expected unqualified-id}} \
                                               // c2x-error {{expected identifier or '('}}
 
 extern int 🌹; // expected-error {{unexpected character <U+1F339>}} \
                   expected-warning {{declaration does not declare anything}}